AMD Ryzen 5 2600 / Ryzen 7 2700 Benchmarks On Linux, 9-Way Ubuntu CPU Comparison

[phoronix]

Phoronix: AMD Ryzen 5 2600 / Ryzen 7 2700 Benchmarks On Linux, 9-Way Ubuntu CPU Comparison

Last month we delivered launch-day Linux benchmarks of the Ryzen 5 2600X and Ryzen 7 2700X for these new "Zen+" processors while recently we received the non-X Ryzen 5 2600 and Ryzen 7 2700 processors for Linux testing as well. In this article are benchmarks of these new AMD Ryzen processors as well as other Intel/AMD CPUs for delivering a fresh nine-way Linux distribution comparison using the very latest software components.

http://www.phoronix.com/vr.php?view=26293

[Jumbotron]

Wow....WOW !! The power consumption stats on the Ryzen 7 2700 are something else !! And on many performance benchmarks it's within the margin of error of the Intel 8700 ! Outstanding !

[humbug]

Looks great. Very tempting to pick up a 2700 for a new build!

also Linux gaming seems to have a problem where our ports are not as well multithreaded as the windows games. These Ryzen+ CPUs perform much closer to Intel in Windows games.

[tildearrow]

Typos:

an upgrade over DDR4-2677 native support with the original Zen CPUs

(are you sure it's exactly 2677?)

while the PB2 boost clock speed is 4.1GHz to 4.3GHz.

(sounds like it goes from 4.1 to 4.3 rather than being 4.1 down from 4.3)

with the i5-8400 running just behind the Core i5 8400,

Haha

while the Ryzeen 5 2600 was

[malakudi]

Michael, your stockfish test is single core, you can test multiple cores usage with following command:
./stockfish bench 128 16 20 default depth
where 16 is the number of cores/threads and 20 is the depth. Higher the depth, better multicore usage.
On my 1700X OC at 3900 MHz and DDR4-3200 CL14, I get:
./stockfish bench 128 1 20 default depth => 68171 ms
./stockfish bench 128 2 20 default depth => 35388 ms
./stockfish bench 128 4 20 default depth => 22008 ms
./stockfish bench 128 8 20 default depth => 20804 ms
./stockfish bench 128 16 20 default depth => 21527 ms

As you can see at depth 20 it doesn't scale well above 4 cores/threads.
At depth 24:
./stockfish bench 128 4 24 default depth => 106362 ms
./stockfish bench 128 8 24 default depth => 82562 ms
./stockfish bench 128 16 24 default depth => 78251 ms

So I would suggest creating a stockfish-multithread test with 24 depth and run at number of threads of each cpu.

The 7zip results seem low, I get 44000 mips on my system. I think Ryzen 2700X should have had higher results if you used DDR4-3400 with correct timings.

[oleyska]

Wow....WOW !! The power consumption stats on the Ryzen 7 2700 are something else !! And on many performance benchmarks it's within the margin of error of the Intel 8700 ! Outstanding !

AMD have truly changed the cpu market, back from the dead for sure!
2700X is a bit effeciency out the window and give me all ya got and it does it really really well dynamically and It shocked me how "on the edge" it is at all times.
2700 is truly effeciency redefined.
2600 is value redefined.

I went into 2018 with expectation of 10% and 15% would impress me, amd gave me expected and promised 10% and I was like ohh, okey.
2700X with it's tech for dynamic frequency and clocking really impressed me and the price points!
Game performance due to reduced latency was more towards 15% at times and it just continues to impress me.

I think Intel's core is better but they've failed to deliver better products, they have the chips but not the products apart from a mighty 8700K... the rest feels very mediocre to me.

[jrch2k8]

https://www.youtube.com/watch?v=PHBsR1Y68G8 michael check that video it may provide a nice boost for ryzen on Linux as well

[schmidtbag]

I find it interesting how the 2400G, despite the extra costs of the IGP, still manages to pull ahead the 8700K in some performance-per-dollar tests.

I think Intel's core is better but they've failed to deliver better products, they have the chips but not the products apart from a mighty 8700K... the rest feels very mediocre to me.

I disagree. There are currently only 3 things that makes Intel's chips better:
1. Lower latencies
2. Higher clocks
3. Better AVX support
In terms of IPC, PPW, and PP$, Ryzen tends to perform better (at least below 3.8GHz). Remember, this is accounting for worse latencies. So, I would argue that Ryzen cores are superior on a functionality standpoint. Intel's cores can do more work in less time, but that doesn't necessarily mean they're better. A 1980's V8 might have more power than the 4-cylinder engine in my car, but my car's engine is still better in a practical sense.

[gsedej]

A bit off-topic about temperature...

My Ryzen 5 1600x has a bit strange graph. When idling, the temperature constantly jumps between 40°C and 50°C. Did anyone else noticed it?
Here is graph with idle and stress test between. Sensor is from AMD k10temp-pci-00c3.

https://imgur.com/a/GB8VT51

[leipero]

There are memoty timings/frequency settings circulating around the web, with such setup, it shows significant improvement in single threaded workloads, it would be interesting to see if such improvements are measurable on GNU/Linux (ofc. if possible here, since it does require some timings that might not work with all RAM modules).